U.S. plans an AI incident channel with China

Treasury Secretary Scott Bessent told Axios that the United States plans to propose an AI incident notification process with China. He said he and Chinese Vice Premier He Lifeng discussed a channel for AI problems, including uncontrolled agents and cyber or biological threats, ahead of President Xi Jinping's state visit. Bessent said U.S. model reviews remain voluntary.

The proposal treats frontier AI as an operational risk that may need communication between competing governments, not only a product-policy issue. Developers serving critical systems could eventually need incident reporting, resilience plans or evidence that agents can be contained.

There is no agreement yet. Watch whether the proposed channel gets a scope, contact point or response procedure, and keep building incident logs that can support a clear account of what an agent did, when it did it and which controls failed.